NCAA Tournament Statistical Projections

[DSMok1]

Anybody up for some stats?

I am using a new algorithm this year, not just the Pomeroy Ratings. A good guide to stats and the NCAA tournament is available at Wired Playbook.

The method I am using looks at 2 more variables (besides simply adjusting for opposition efficiency like Pomeroy). I ran a regression using the game number and the opposition strength as predictors. This allows us to see "hot" and "cold" teams, and teams that "beat up on the little teams" vs. teams that "play to the level of their competition.

What do these numbers mean? Well, for starters, these are all efficiency differentials. So Syracuse, for instance, started this season as a +33.8 efficiency (points/100 Poss) team. Every game, they have dropped 0.16 in efficiency--they are somewhat "cold". However, the last number is how much they "play to the level of their competition". This number is how much better they would play against an extremely good team. That helps them a LOT in this tournament.

Here are the 64 teams through that lens:

Team		BaseMar	Season	OppVar
Syracuse	33.8	-0.16	9.2
Duke		36.2	-0.09	0.4
Kansas		40.5	-0.37	3.2
Baylor		25.9	0.12	4.3
Northern Iowa	16.1	0.43	4.9
Gonzaga		8.4	0.56	13.5
Kansas St.	26.6	-0.01	5.6
Ohio St.	27.0	-0.06	8.6
Georgetown	20.5	0.37	-2.0
Maryland	15.3	0.76	-8.2
Cornell		8.1	0.43	11.0
Kentucky	20.9	0.50	-7.5
Butler		12.9	0.35	7.1
Wisconsin	27.4	0.20	-6.8
Purdue		27.6	-0.09	2.3
West Virginia	27.5	-0.09	-0.3
Pittsburgh	8.0	0.80	-2.7
Temple		24.3	-0.05	2.2
Washington	8.7	0.58	3.4
Oklahoma St.	8.5	0.67	-0.3
Clemson		16.9	0.50	-8.0
Marquette	28.7	-0.24	1.6
Richmond	15.2	0.14	5.5
New Mexico	24.9	-0.35	8.7
San Diego St.	4.8	0.78	-1.9
Notre Dame	17.0	0.01	8.4
Villanova	27.0	-0.16	-2.2
Missouri	25.3	-0.19	2.8
Utah St.	13.0	0.58	-6.4
Nevada Las Vega	18.5	0.07	2.6
Texas A&M	22.1	-0.06	2.1
California	27.0	-0.04	-6.3
Florida St.	25.3	0.07	-10.0
Georgia Tech	20.8	-0.05	2.5
Wake Forest	21.2	-0.12	2.8
Florida		20.0	-0.19	7.6
Louisville	20.0	0.01	-1.4
Brigham Young	23.2	0.03	-5.9
Vanderbilt	23.1	-0.22	2.0
Xavier		21.1	0.09	-6.7
Minnesota	20.4	0.08	-8.1
Texas El Paso	9.9	0.35	-0.1
Texas		37.0	-0.71	-3.0
Michigan St.	20.5	0.09	-8.9
Tennessee	22.5	-0.24	0.8
Murray St.	13.4	0.10	-0.5
Vermont		-5.3	0.58	4.1
St. Mary's	22.3	-0.34	-0.3
Houston		2.4	0.37	2.3
Montana		9.3	0.01	3.9
Ohio		2.4	0.25	4.9
Old Dominion	24.1	-0.16	-9.3
New Mexico St.	-5.0	0.63	2.7
Siena		16.5	-0.19	-1.2
East Tennessee 	2.3	0.24	1.7
Lehigh		-14.3	0.79	4.5
Wofford		-2.3	0.43	0.2
UC Santa Barbar	-6.2	0.43	4.2
Sam Houston St.	18.7	-0.56	-2.7
Morgan St.	6.9	-0.18	3.7
North Texas	-0.2	0.08	3.3
Robert Morris	-13.9	0.60	0.8
Oakland		0.7	0.06	-5.1
Arkansas PB	-9.7	0.02	7.3

Here's a list of the hottest teams:

Team		BaseMar	Season	OppVar
Pittsburgh	8.0	0.80	-2.7
Lehigh		-14.3	0.79	4.5
San Diego St.	4.8	0.78	-1.9
Maryland	15.3	0.76	-8.2
Oklahoma St.	8.5	0.67	-0.3
New Mexico St.	-5.0	0.63	2.7
Robert Morris	-13.9	0.60	0.8
Vermont		-5.3	0.58	4.1
Washington	8.7	0.58	3.4
Utah St.	13.0	0.58	-6.4
Gonzaga		8.4	0.56	13.5
Kentucky	20.9	0.50	-7.5
Clemson		16.9	0.50	-8.0
Wofford		-2.3	0.43	0.2
Cornell		8.1	0.43	11.0

Of course, most of the hottest teams aren't very good--they won their conference tournaments to get in.

There are a few to note, however--Maryland is really hot. So is Pittsburgh. OSU is hot, and Gonzaga and Kentucky are hot.

The cold teams:

Team		BaseMar	Season	OppVar
Texas		37.0	-0.71	-3.0
Sam Houston St.	18.7	-0.56	-2.7
Kansas		40.5	-0.37	3.2
New Mexico	24.9	-0.35	8.7
St. Mary's	22.3	-0.34	-0.3
Tennessee	22.5	-0.24	0.8
Marquette	28.7	-0.24	1.6
Vanderbilt	23.1	-0.22	2.0
Florida		20.0	-0.19	7.6
Siena		16.5	-0.19	-1.2
Missouri	25.3	-0.19	2.8
Morgan St.	6.9	-0.18	3.7
Old Dominion	24.1	-0.16	-9.3
Syracuse	33.8	-0.16	9.2
Villanova	27.0	-0.16	-2.2

Obviously, Texas is the coldest. The started the season REALLY hot. Interestingly, Kansas shows up as fading somewhat also... other than Texas, though, the cold teams aren't as cold as the hot teams are hot.

More coming in the next post...

 

[DSMok1]

On to the other number: OppVar. This looks at how much the team "played to the level of their opposition." Some teams demolished the bad teams with pure talent but struggled against good, disciplined teams. Kentucky is the prime example. Others just played to win against the bad teams and didn't run up the score at all, but turned up the heat against the good teams. That would be Gonzaga, Syracuse, and Cornell.

Obviously, this cannot be a linear relationship like the previous number. I used an Error Function as the base for this model. The result of the error function varies from -1 to 1; I used it on the z-scores of opponents (stdevs above and below average opponents) to create a custom prediction for each team. Some teams didn't vary much with opponents; others did a lot. The number listed here is the change if playing an "infinitely good" team--actually the modification depends on how many standard deviations the team they are playing is above the average opponent they played during the year.

(I know, scary... but it's a good way to look at it).

So anyway: here are the teams that "raise their game" against the good teams:

Team		BaseMar	Season	OppVar	AvgOpp
Gonzaga		8.4	0.56	13.5	4.3
Cornell		8.1	0.43	11.0	-3.4
Syracuse	33.8	-0.16	9.2	11.2
New Mexico	24.9	-0.35	8.7	5.8
Ohio St.	27.0	-0.06	8.6	9.8
Notre Dame	17.0	0.01	8.4	9.2
Florida		20.0	-0.19	7.6	8.8
Arkansas PB	-9.7	0.02	7.3	-8.9
Butler		12.9	0.35	7.1	5.3
Kansas St.	26.6	-0.01	5.6	13.8
Richmond	15.2	0.14	5.5	6.2
Ohio		2.4	0.25	4.9	0.6
Northern Iowa	16.1	0.43	4.9	3.7
Lehigh		-14.3	0.79	4.5	-9.8
Baylor		25.9	0.12	4.3	10.8

Gonzaga sleep-walked through the WCC (AGAIN!)... they do that every year. But line them up against a top team and they can play with them. So they are dangerous. Cornell and Syracuse are too... Cornell played mostly bad teams, so they will be getting the full 11 point bonus against most teams in the tournament. They played Syracuse tough and just about beat Kansas...both on the road. Watch out for them. Syracuse coasted in the non-conf and then obliterated most of the Big East.

And the teams that faltered against good teams but stomped the little guys to pad their numbers:

Team		BaseMar	Season	OppVar	AvgOpp
Florida St.	25.3	0.07	-10.0	11.1
Old Dominion	24.1	-0.16	-9.3	3.6
Michigan St.	20.5	0.09	-8.9	10.5
Maryland	15.3	0.76	-8.2	11.3
Minnesota	20.4	0.08	-8.1	10.8
Clemson		16.9	0.50	-8.0	11.2
Kentucky	20.9	0.50	-7.5	9.2
Wisconsin	27.4	0.20	-6.8	11.6
Xavier		21.1	0.09	-6.7	8.1
Utah St.	13.0	0.58	-6.4	2.1
California	27.0	-0.04	-6.3	10.8
Brigham Young	23.2	0.03	-5.9	3.8
Oakland		0.7	0.06	-5.1	-2.0
Texas		37.0	-0.71	-3.0	10.7
Sam Houston St.	18.7	-0.56	-2.7	-3.5

Florida State is likely fool's gold. Against the bad teams, they excelled. Against the good teams... they were just mediocre. A total of a 20 point swing! Old Dominion, Michigan State, and Maryland all have similar issues. (Maryland is also hot, though, so watch for them anyway).

Remember--most of the teams in the tournament are good (~+19 margin). If a team has struggled against good teams, that's a really bad sign.

In the next post, I'll combine the two and show how teams could likely do.

 

[DSMok1]

So now... looking at good teams that are hot AND play good against good teams.

First of all.. let's mitigate the "hot" factor somewhat. We can't expect teams to just keep getting hotter. I'll assume each team plays as if it is 75% of the way through the season.

Next, applying the adjustment for opposition. Unfortunately, that depends on the opponent. When running the tourney predictor, I'll look at each match up separately. But here, I want to get an overall estimate of how strong the teams are.

The average tournament team is somewhere around a +19 margin. I'll assume each opponent is a +20 to get a rough idea of how strong teams will be in the tournament.

Seed	Team		InitMar	YearMar	EstEff	StDev
1	Syracuse	33.8	30.0	36.1	13.5
1	Duke		36.2	33.7	34.0	13.7
8	Gonzaga		8.4	21.8	33.7	18.4
1	Kansas		40.5	31.2	33.2	13.1
3	Baylor		25.9	28.8	31.5	14.5
9	Northern Iowa	16.1	26.3	31.1	13.2
2	Ohio St.	27.0	25.4	30.9	17.9
2	Kansas St.	26.6	26.4	29.3	12.9
3	Georgetown	20.5	29.7	28.8	14.8
12	Cornell		8.1	18.1	28.7	14.1
4	Maryland	15.3	33.0	28.2	14.0
5	Butler		12.9	21.2	27.9	12.4
4	Wisconsin	27.4	32.0	27.9	18.4
1	Kentucky	20.9	33.6	27.0	12.4
4	Purdue		27.6	25.4	26.9	13.5
11	Washington	8.7	23.1	26.2	20.2
3	Pittsburgh	8.0	27.3	25.3	15.6
2	West Virginia	27.5	25.4	25.2	12.3
5	Temple		24.3	23.1	24.9	14.5
7	Oklahoma St.	8.5	24.7	24.5	14.6
3	New Mexico	24.9	16.1	24.0	13.8
6	Marquette	28.7	22.8	23.8	12.7
6	Notre Dame	17.0	17.3	23.6	13.7
7	Clemson		16.9	28.4	23.6	12.8
7	Richmond	15.2	18.6	23.5	13.8
10	Missouri	25.3	20.8	22.7	16.3
11	San Diego St.	4.8	24.2	22.4	13.8
8	Nevada Las Vegas18.5	20.2	22.4	15.6
2	Villanova	27.0	23.4	22.1	12.7
5	Texas A&M	22.1	20.7	21.8	9.5
12	Utah St.	13.0	27.8	21.5	15.9
9	Florida St.	25.3	26.9	21.0	14.4
8	California	27.0	26.0	21.0	14.6
10	Georgia Tech	20.8	19.6	21.0	11.0
10	Florida		20.0	15.3	20.7	12.2
9	Wake Forest	21.2	18.6	20.3	13.3
4	Vanderbilt	23.1	17.9	19.6	15.7
9	Louisville	20.0	20.2	19.2	18.5
7	Brigham Young	23.2	24.1	18.7	16.3
12	Texas El Paso	9.9	18.4	18.3	14.5
6	Xavier		21.1	23.3	17.7	12.0
8	Texas		37.0	19.4	17.4	13.2
11	Minnesota	20.4	22.5	17.2	19.9
6	Tennessee	22.5	16.5	17.1	18.1
5	Michigan St.	20.5	22.7	16.5	14.7
13	Murray St.	13.4	15.9	15.3	14.1
13	Houston		2.4	11.9	14.1	13.2
10	St. Mary's	22.3	14.4	14.1	13.7
14	Ohio		2.4	9.0	13.9	12.2
16	Vermont		-5.3	9.6	13.7	17.6
14	Montana		9.3	9.6	13.5	15.5
12	New Mexico St.	-5.0	10.6	13.0	14.2
11	Old Dominion	24.1	19.9	11.1	15.5
13	Siena		16.5	11.8	10.7	11.2
16	East Tennessee S2.3	8.4	10.1	14.5
16	Lehigh		-14.3	4.8	9.3	14.1
13	Wofford		-2.3	8.7	8.9	10.6
15	UC Santa Barbar	-6.2	3.1	7.3	13.7
15	Morgan St.	6.9	2.1	5.8	12.4
15	North Texas	-0.2	1.8	5.1	11.3
14	Sam Houston St.	18.7	5.2	2.5	18.2
15	Robert Morris	-13.9	1.4	2.2	11.8
16	Arkansas PB	-9.7	-9.3	-2.2	9.5
14	Oakland		0.7	2.3	-2.6	14.8
17	Winthrop	-13.1	-8.0	-5.8	14.4

The teams that are hot AND play well against good teams rise to the top.

However, there is another consideration--how schizo the teams are. Even if you are really good, hot, and play well against the good teams, if you MAIL IT IN every once in a while that will bite you sooner or later. Gonzaga did against Duke.

Thus, when actually calculating the odds of who will win a game, we must look at the variability of the team. That would be the StDev (standard deviation) listed in the table above. That StDev is how accurately my model accounted for all of the games. Gonzaga... was schizo. Kansas rarely mailed it in.

So we want a hot team, that plays well against the big boys, that is consistent.

OR, if the team is an underdog, one that is INconsistent. If the team is of lower caliber, we want the team to be able to jump up and bite good teams sometimes. A Tennessee, anyone?

Finally the bracket in the next post...

 

[DSMok1]

And here is the bracket:

	Rank	Seed	Team		VsTourn	32	16	8	4	2	Champ
Midwest	3	1	Kansas		0.7374	86.6%	53.0%	38.8%	24.5%	13.6%	8.4%
	56	16	Lehigh		0.3235	13.4%	2.7%	0.5%	0.1%	0.0%	0.0%
	30	8	Nevada Las Vegas0.5451	34.0%	12.0%	5.9%	2.6%	0.9%	0.4%
	5	9	Northern Iowa	0.6982	66.0%	32.3%	22.2%	12.9%	6.5%	3.6%
	44	5	Michigan St.	0.4688	65.3%	19.3%	4.3%	1.3%	0.3%	0.1%
	53	12	New Mexico St.	0.3783	34.7%	8.0%	1.7%	0.5%	0.1%	0.0%
	10	4	Maryland	0.6531	81.3%	64.6%	24.9%	12.3%	5.0%	2.3%
	49	13	Houston		0.3937	18.7%	8.0%	1.7%	0.5%	0.1%	0.0%

	45	6	Tennessee	0.4598	40.5%	14.4%	5.3%	1.6%	0.5%	0.2%
	25	11	San Diego St.	0.5582	59.5%	26.3%	11.3%	4.1%	1.5%	0.6%
	9	3	Georgetown	0.6599	79.5%	51.7%	26.9%	12.8%	5.7%	2.9%
	51	14	Ohio		0.3848	20.5%	7.6%	2.3%	0.5%	0.1%	0.0%
	20	7	Oklahoma St.	0.5891	56.5%	26.4%	13.4%	5.4%	2.1%	0.9%
	34	10	Georgia Tech	0.5252	43.5%	18.5%	8.1%	2.7%	0.9%	0.3%
	8	2	Ohio St.	0.6698	76.8%	50.0%	31.7%	18.0%	9.6%	5.5%
	58	15	UC Santa Barbara0.2848	23.2%	5.1%	1.0%	0.1%	0.0%	0.0%

	Rank	Seed	Team		VsTourn	32	16	8	4	2	Champ
West	1	1	Syracuse	0.7724	79.5%	50.1%	37.9%	27.4%	18.5%	12.8%
	47	16	Vermont		0.4048	20.5%	5.5%	2.0%	0.6%	0.2%	0.0%
	6	8	Gonzaga		0.6956	76.3%	38.0%	25.7%	16.6%	9.9%	6.0%
	33	9	Florida St.	0.5380	23.7%	6.4%	2.9%	1.1%	0.3%	0.1%
	13	5	Butler		0.6407	68.9%	47.4%	17.7%	9.8%	4.7%	2.4%
	42	12	Texas El Paso	0.4790	31.1%	16.0%	4.1%	1.6%	0.5%	0.2%
	39	4	Vanderbilt	0.4937	56.7%	22.7%	6.4%	2.6%	0.9%	0.3%
	46	13	Murray St.	0.4263	43.3%	14.0%	3.3%	1.1%	0.3%	0.1%

	40	6	Xavier		0.4836	50.3%	18.5%	6.7%	1.7%	0.5%	0.1%
	41	11	Minnesota	0.4820	49.7%	18.8%	7.1%	2.1%	0.6%	0.2%
	17	3	Pittsburgh	0.6052	93.4%	62.2%	30.0%	11.4%	4.8%	2.1%
	63	14	Oakland		0.1915	6.6%	0.5%	0.1%	0.0%	0.0%	0.0%
	38	7	Brigham Young	0.4996	46.9%	18.2%	7.9%	2.4%	0.7%	0.2%
	36	10	Florida		0.5042	53.1%	21.1%	10.1%	3.4%	1.2%	0.5%
	7	2	Kansas St.	0.6732	84.4%	56.8%	37.6%	18.1%	9.8%	5.5%
	61	15	North Texas	0.2450	15.6%	3.9%	0.5%	0.1%	0.0%	0.0%

	Rank	Seed	Team		VsTourn	32	16	8	4	2	Champ
East	12	1	Kentucky	0.6512	88.4%	59.5%	29.4%	16.2%	7.2%	2.9%
	55	16	East Tennessee 	0.3363	11.6%	3.8%	0.8%	0.2%	0.0%	0.0%
	43	8	Texas		0.4690	44.6%	14.1%	4.3%	1.6%	0.4%	0.1%
	35	9	Wake Forest	0.5102	55.4%	22.6%	8.8%	3.8%	1.3%	0.4%
	18	5	Temple		0.5920	42.6%	22.0%	11.8%	6.2%	2.6%	1.0%
	11	12	Cornell		0.6525	57.4%	33.9%	20.8%	12.4%	6.2%	2.8%
	14	4	Wisconsin	0.6362	87.7%	41.8%	23.7%	12.9%	5.8%	2.4%
	57	13	Wofford		0.3013	12.3%	2.3%	0.4%	0.1%	0.0%	0.0%
	22	6	Marquette	0.5785	46.3%	25.8%	13.1%	6.1%	2.5%	0.9%

	19	11	Washington	0.5895	53.7%	31.0%	17.0%	8.7%	4.0%	1.7%
	24	3	New Mexico	0.5587	65.3%	32.0%	16.6%	7.9%	3.3%	1.2%
	50	14	Montana		0.3930	34.7%	11.2%	3.3%	0.9%	0.2%	0.1%
	21	7	Clemson		0.5823	49.1%	24.1%	11.5%	4.9%	1.7%	0.5%
	28	10	Missouri	0.5505	50.9%	25.5%	12.5%	5.6%	2.3%	0.8%
	16	2	West Virginia	0.6074	87.1%	48.3%	25.5%	12.4%	5.2%	2.0%
	60	15	Morgan St.	0.2606	12.9%	2.1%	0.5%	0.1%	0.0%	0.0%

	Rank	Seed	Team		VsTourn	32	16	8	4	2	Champ
South	2	1	Duke		0.7458	98.3%	74.1%	51.4%	33.2%	22.3%	12.2%
	64	16	Arkansas PB	0.1409	1.7%	0.1%	0.0%	0.0%	0.0%	0.0%
	32	8	California	0.5441	53.6%	13.4%	5.9%	2.4%	0.9%	0.3%
	37	9	Louisville	0.5004	46.4%	12.3%	5.2%	2.1%	0.8%	0.3%
	31	5	Texas A&M	0.5445	51.0%	23.5%	7.6%	3.2%	1.4%	0.4%
	29	12	Utah St.	0.5458	49.0%	22.6%	7.7%	3.3%	1.4%	0.5%
	15	4	Purdue		0.6319	80.0%	48.8%	21.5%	11.2%	6.1%	2.6%
	54	13	Siena		0.3369	20.0%	5.2%	0.7%	0.2%	0.0%	0.0%

	26	6	Notre Dame	0.5543	64.8%	27.0%	15.0%	6.3%	3.1%	1.2%
	52	11	Old Dominion	0.3835	35.2%	10.0%	3.2%	0.6%	0.2%	0.0%
	4	3	Baylor		0.7022	84.6%	58.5%	41.1%	23.3%	14.8%	7.7%
	59	14	Sam Houston St.	0.2607	15.4%	4.4%	1.0%	0.1%	0.0%	0.0%
	23	7	Richmond	0.5592	66.0%	37.4%	16.2%	6.5%	3.1%	1.1%
	48	10	St. Mary's	0.4040	34.0%	12.2%	3.8%	0.9%	0.2%	0.1%
	27	2	Villanova	0.5522	90.5%	48.3%	19.4%	6.8%	2.9%	0.9%
	62	15	Robert Morris	0.2124	9.5%	2.2%	0.4%	0.0%	0.0%	0.0%

Potential surprise teams:
-- Northern Iowa? They're hot, good against good teams, AND consistent. Watch out for them.
-- Cornell is dangerous
-- Syracuse is a slight favorite over Duke
-- Villanova ain't going far
-- Baylor is really dangerous. (I NEVER thought I'd say that)

Strength of the brackets by likelihood of winning the title:
Midwest 25%
West 30%
East 17%
South 27%

5 teams from the West can make some noise.

Enjoy, and pick a good bracket folks! The Wired News link at the top gives a good idea how to take advantage of the numerical analysis. Don't go with the crowds... pick underrated teams.

 

[sooner333]

Don't know if I trust something that says Kansas isn't even 90% to beat Lehigh, but I do enjoy the statistical analysis!

 

[thebigabd]

Is this going to be on the test?

 

[DSMok1]

Don't know if I trust something that says Kansas isn't even 90% to beat Lehigh, but I do enjoy the statistical analysis!

Yeah, I don't really buy that either...

But then again, when looking at how each team played: Lehigh's best 5 performances would beat Kansas's 5 worst performances. So it is possible. Here are the ADJUSTED efficiency margins for each game each team played, from best to worst. Also, Kansas isn't hot and doesn't play as dominantly against poor teams.. while Lehigh is VERY hot and plays best against good teams. I have Lehigh as 10th worst in the tournament, by no means a 16 seed in quality.

Kansas	Lehigh
Margin	Margin
75.7	25.1
57.9	23.0
54.9	20.1
48.8	17.1
47.7	14.2
45.9	11.6
44.7	10.9
42.6	7.4
42.3	6.8
39.4	6.5
39.4	5.2
39.2	5.1
39.1	3.5
38.5	2.4
37.4	2.0
35.9	1.7
35.3	0.7
35.2	-1.0
34.6	-3.4
32.3	-4.7
30.5	-7.4
30.4	-9.5
30.2	-10.0
30.0	-12.9
27.4	-15.0
24.9	-16.7
24.4	-18.5
23.3	-21.4
21.5	-22.3
21.2	-24.6
20.6	-28.6
15.8	-35.8
13.1	
12.6

 

[DSMok1]

Is this going to be on the test?

This is original research as far as I know, so no.

Then again, in Advanced Finite Element Analysis, my professor had us re-derive his doctoral dissertation's key innovation (behavior of partially-saturated soil). Ouch.

 

[NickZepp]

Does this factor in injuries? For example Purdue and Missouri are not going to have some of their top players.

 

[DSMok1]

Does this factor in injuries? For example Purdue and Missouri are not going to have some of their top players.

Nope, it doesn't... but that can be reflected in the fact they are getting worse at the end of the year.

That said,

I MADE A BIG MISTAKE on this... I failed to adjust for the significance of the regression. Basically, you can have a steep slope on your best-fit line, but that doesn't mean that the best fit line is really much better than a flat line at matching the data. You can pick up a trend in random data... that doesn't mean it will continue. I have to adjust for that fact.

So I've got to adjust. Basically, the trends/factors do not have as big an impact as I showed. I'm going to weight it according to the "explanatory power" of the regression, the r^2.

 

[NickZepp]

Purdue was playing really good and even beat Ohio State near the end of the year before Hummell got hurt. But they still won 8 of their last 10.

 

[DSMok1]

Sorry for that BIG mistake!

Here are the corrected projections... this looks a lot more like, say, the Pomeroy projections now.

	Rank	Seed	Team		VsTourn	32	16	8	4	2	Champ
Midwest	2	1	Kansas		0.785	95.5%	74.4%	54.1%	37.9%	26.3%	16.7%
	61	16	Lehigh		0.232	4.5%	0.8%	0.1%	0.0%	0.0%	0.0%
	41	8	Nevada Las Vegas0.511	45.4%	10.7%	4.5%	1.8%	0.7%	0.2%
	28	9	Northern Iowa	0.552	54.6%	14.1%	6.5%	2.9%	1.2%	0.4%
	20	5	Michigan St.	0.583	75.2%	36.6%	12.3%	5.8%	2.7%	1.1%
	52	12	New Mexico St.	0.338	24.8%	6.0%	0.9%	0.2%	0.0%	0.0%
	10	4	Maryland	0.650	80.1%	50.8%	20.6%	11.2%	5.9%	2.8%
	50	13	Houston		0.365	19.9%	6.7%	1.0%	0.3%	0.1%	0.0%

	32	6	Tennessee	0.538	52.2%	24.0%	10.5%	3.7%	1.5%	0.6%
	39	11	San Diego St.	0.520	47.8%	20.6%	8.4%	2.5%	1.0%	0.3%
	11	3	Georgetown	0.638	81.4%	50.0%	26.8%	11.6%	6.0%	2.8%
	55	14	Ohio		0.328	18.6%	5.4%	1.2%	0.2%	0.0%	0.0%
	43	7	Oklahoma St.	0.502	45.3%	17.1%	7.2%	2.1%	0.8%	0.2%
	31	10	Georgia Tech	0.543	54.7%	22.1%	10.1%	3.1%	1.2%	0.4%
	7	2	Ohio St.	0.666	88.5%	58.7%	35.3%	16.7%	9.3%	4.7%
	59	15	UC Santa Barbara0.238	11.5%	2.1%	0.4%	0.0%	0.0%	0.0%

West	3	1	Syracuse	0.706	88.1%	59.5%	41.3%	25.1%	13.1%	7.0%
	56	16	Vermont		0.280	11.9%	2.8%	0.7%	0.1%	0.0%	0.0%
	42	8	Gonzaga		0.503	43.1%	15.1%	7.1%	2.8%	0.9%	0.3%
	22	9	Florida St.	0.574	56.9%	22.7%	12.4%	5.6%	2.1%	0.8%
	25	5	Butler		0.571	55.4%	31.7%	13.3%	6.0%	2.2%	0.8%
	37	12	Texas El Paso	0.523	44.6%	23.1%	8.7%	3.5%	1.2%	0.4%
	33	4	Vanderbilt	0.535	57.4%	27.9%	11.0%	4.7%	1.6%	0.6%
	48	13	Murray St.	0.466	42.6%	17.3%	5.5%	1.9%	0.5%	0.1%

	21	6	Xavier		0.578	49.7%	29.6%	12.4%	5.6%	2.1%	0.8%
	27	11	Minnesota	0.564	50.3%	29.1%	12.8%	6.0%	2.4%	1.0%
	30	3	Pittsburgh	0.545	79.0%	37.4%	15.0%	6.5%	2.3%	0.8%
	57	14	Oakland		0.265	21.0%	4.0%	0.6%	0.1%	0.0%	0.0%
	9	7	Brigham Young	0.653	69.7%	37.1%	22.7%	12.6%	6.0%	2.9%
	47	10	Florida		0.474	30.3%	9.7%	4.0%	1.4%	0.4%	0.1%
	8	2	Kansas St.	0.661	93.5%	52.5%	32.4%	18.0%	8.4%	4.0%
	62	15	North Texas	0.201	6.5%	0.7%	0.1%	0.0%	0.0%	0.0%

	Rank	Seed	Team		VsTourn	32	16	8	4	2	Champ
East	4	1	Kentucky	0.679	91.6%	60.7%	34.5%	20.0%	9.8%	4.8%
	58	16	East Tennessee 	0.249	8.4%	1.6%	0.3%	0.0%	0.0%	0.0%
	26	8	Texas		0.570	59.2%	24.4%	11.2%	5.1%	1.9%	0.7%
	44	9	Wake Forest	0.489	40.8%	13.3%	5.1%	1.9%	0.6%	0.2%
	23	5	Temple		0.573	58.7%	26.4%	11.8%	5.5%	2.1%	0.8%
	45	12	Cornell		0.487	41.3%	15.5%	5.6%	2.1%	0.7%	0.2%
	6	4	Wisconsin	0.671	82.5%	53.3%	30.7%	18.0%	9.2%	4.8%
	54	13	Wofford		0.334	17.5%	4.9%	0.9%	0.2%	0.0%	0.0%

	19	6	Marquette	0.583	52.6%	32.9%	14.3%	6.3%	2.4%	1.0%
	29	11	Washington	0.545	47.4%	27.3%	11.6%	5.0%	1.9%	0.7%
	38	3	New Mexico	0.521	69.5%	31.5%	11.8%	4.6%	1.5%	0.5%
	53	14	Montana		0.335	30.5%	8.2%	1.6%	0.3%	0.1%	0.0%
	16	7	Clemson		0.600	50.4%	22.1%	12.5%	5.8%	2.3%	0.9%
	18	10	Missouri	0.587	49.6%	22.1%	12.3%	5.6%	2.3%	1.0%
	5	2	West Virginia	0.676	92.6%	54.8%	35.6%	19.4%	9.4%	4.6%
	60	15	Morgan St.	0.236	7.4%	1.0%	0.2%	0.0%	0.0%	0.0%

South	1	1	Duke		0.792	99.3%	72.0%	53.8%	40.1%	28.0%	18.1%
	64	16	Arkansas PB	0.113	0.7%	0.0%	0.0%	0.0%	0.0%	0.0%
	15	8	California	0.623	58.6%	17.6%	9.6%	5.2%	2.6%	1.1%
	35	9	Louisville	0.531	41.4%	10.3%	4.6%	2.1%	0.9%	0.3%
	24	5	Texas A&M	0.572	46.0%	22.7%	6.2%	2.9%	1.2%	0.4%
	17	12	Utah St.	0.594	54.0%	28.5%	9.9%	5.1%	2.4%	1.0%
	14	4	Purdue		0.625	71.9%	39.7%	14.2%	7.8%	3.9%	1.7%
	49	13	Siena		0.436	28.1%	9.2%	1.6%	0.5%	0.2%	0.0%

	40	6	Notre Dame	0.517	49.5%	22.1%	9.7%	3.0%	1.1%	0.4%
	36	11	Old Dominion	0.526	50.5%	22.9%	10.2%	3.2%	1.2%	0.4%
	12	3	Baylor		0.633	77.7%	48.1%	27.5%	11.6%	6.0%	2.8%
	51	14	Sam Houston St.	0.339	22.3%	7.0%	1.8%	0.3%	0.1%	0.0%
	46	7	Richmond	0.481	44.5%	16.7%	6.9%	1.9%	0.6%	0.2%
	34	10	St. Mary's	0.533	55.5%	23.8%	11.0%	3.4%	1.4%	0.5%
	13	2	Villanova	0.625	93.1%	58.5%	32.7%	12.8%	6.3%	2.7%
	63	15	Robert Morris	0.187	6.9%	1.0%	0.1%	0.0%	0.0%	0.0%

For most of the teams, all of the math I did didn't really effect them... assuming they are the same as they were all season was just as accurate as the "hot" or "not." But for some, it is still important:

Team		Hotness
New Mexico St.	0.16
Lehigh		0.13
Robert Morris	0.08
Cornell		0.08
Utah St.	0.08
Maryland	0.08
Kentucky	0.06
San Diego St.	0.06
Northern Iowa	0.05
Pittsburgh	0.03

(Hotness is in points improving, per game)
Cold Teams:

Team		Hotness
Texas		-0.15
New Mexico	-0.04
St. Mary's	-0.01
Kansas		-0.01
Tennessee	-0.01

Just Texas, basically.
Teams strong against the best:

Team		To Level of Opp
Cornell		2.11
Arkansas PB	1.35
Notre Dame	1.16
New Mexico	0.88
Lehigh		0.75
New Mexico St.	0.69
Northern Iowa	0.59
Baylor		0.54
Ohio		0.53
Butler		0.52
Richmond	0.52
Gonzaga		0.51
Syracuse	0.33
North Texas	0.29
Florida		0.25

Weak against the best:

Team		To Level of Opp
Xavier		-1.32
Utah St.	-0.85
Kentucky	-0.84
Maryland	-0.81
Texas		-0.66
Old Dominion	-0.48
Oakland		-0.46
Florida St.	-0.41
Minnesota	-0.36
Clemson		-0.27

I'm sorry about that! That was a basic statistics error.

 

[DSMok1]

Okay, here are the VALUE PICKS.

That means the picks that have the greatest difference between the unbiased statistics and the average bracket (as measured by ESPN.com).

This assumes the rounds are weighted 1,2,4,8,16,32 like at ESPN.com.

Team		Seed	Region	32 Val	16 Val	8 Val	4 Val	2 Val	Champ Val
Kansas		1	MW	-0.04	-0.52	-1.96	-4.75	-9.60	-16.7
Lehigh		16	MW	0.04	0.05	0.05	0.03	0.02	0.0
Nevada Las Vega	8	MW	-0.06	0.14	0.31	0.44	0.54	0.6
Northern Iowa	9	MW	0.06	0.32	0.57	0.79	0.97	1.1
Michigan St.	5	MW	-0.15	-0.45	-0.15	0.15	0.40	0.5
New Mexico St.	12	MW	0.15	0.23	0.26	0.28	0.28	0.3
Maryland	4	MW	-0.10	0.03	0.70	1.47	2.28	3.0
Houston		13	MW	0.10	0.19	0.22	0.23	0.24	0.2
Tennessee	6	MW	-0.26	-0.19	-0.06	0.13	0.29	0.4
San Diego St.	11	MW	0.26	0.63	0.95	1.14	1.30	1.4
Georgetown	3	MW	-0.16	-0.70	-1.35	-1.12	-0.83	-0.5
Ohio		14	MW	0.16	0.26	0.30	0.32	0.32	0.3
Oklahoma St.	7	MW	-0.02	0.18	0.39	0.54	0.65	0.7
Georgia Tech	10	MW	0.02	0.32	0.66	0.88	1.05	1.2
Ohio St.	2	MW	-0.11	-0.63	-1.03	-0.68	-0.44	-0.2
UC Santa Barbara15	MW	0.11	0.14	0.14	0.14	0.15	0.1

Syracuse	1	W	-0.11	-0.77	-2.46	-4.96	-6.06	-7.5
Vermont		16	W	0.11	0.16	0.18	0.17	0.16	0.2
Gonzaga		8	W	-0.25	-0.06	0.10	0.26	0.37	0.4
Florida St.	9	W	0.25	0.68	1.14	1.58	1.90	2.1
Butler		5	W	-0.17	-0.33	-0.01	0.35	0.68	0.9
Texas El Paso	12	W	0.17	0.43	0.76	1.02	1.19	1.3
Vanderbilt	4	W	-0.27	-0.64	-0.45	-0.23	-0.02	0.1
Murray St.	13	W	0.27	0.53	0.74	0.88	0.96	1.0
Xavier		6	W	-0.14	0.07	0.40	0.78	1.08	1.3
Minnesota	11	W	0.14	0.55	1.01	1.46	1.82	2.1
Pittsburgh	3	W	-0.19	-0.88	-1.38	-1.59	-1.49	-1.4
Oakland		14	W	0.19	0.25	0.27	0.28	0.28	0.3
Brigham Young	7	W	0.07	0.66	1.44	2.38	3.30	4.2
Florida		10	W	-0.07	0.04	0.15	0.24	0.26	0.3
Kansas St.	2	W	-0.06	-0.76	-1.94	-2.66	-2.12	-1.3
North Texas	15	W	0.06	0.06	0.05	0.05	0.05	0.0

Team		Seed	Region	32 Val	16 Val	8 Val	4 Val	2 Val	Champ Val
Kentucky	1	E	-0.08	-0.73	-2.77	-6.15	-12.28	-18.2
East Tennessee 	16	E	0.08	0.10	0.11	0.09	0.08	0.1
Texas		8	E	-0.07	0.31	0.62	0.92	1.13	1.3
Wake Forest	9	E	0.07	0.32	0.50	0.63	0.71	0.8
Temple		5	E	-0.09	-0.16	0.21	0.60	0.90	1.1
Cornell		12	E	0.09	0.17	0.35	0.51	0.60	0.7
Wisconsin	4	E	-0.13	-0.22	0.73	1.97	3.28	4.7
Wofford		13	E	0.13	0.21	0.24	0.26	0.26	0.3
Marquette	6	E	-0.24	-0.42	-0.17	0.24	0.57	0.8
Washington	11	E	0.24	0.62	1.03	1.40	1.68	1.9
New Mexico	3	E	-0.26	-0.61	-0.65	-0.50	-0.42	-0.3
Montana		14	E	0.26	0.41	0.47	0.50	0.51	0.5
Clemson		7	E	-0.03	0.33	0.77	1.21	1.56	1.9
Missouri	10	E	0.03	0.41	0.85	1.29	1.64	1.9
West Virginia	2	E	-0.06	-0.82	-2.38	-3.04	-4.25	-4.2
Morgan St.	15	E	0.06	0.08	0.07	0.07	0.07	0.1

Duke		1	S	0.00	-0.26	-1.06	-1.35	0.36	4.0
Arkansas PB	16	S	0.00	0.00	-0.01	-0.02	-0.04	-0.1
California	8	S	0.31	0.62	0.96	1.35	1.75	2.1
Louisville	9	S	-0.31	-0.36	-0.51	-0.52	-0.48	-0.4
Texas A&M	5	S	-0.33	-0.75	-0.79	-0.67	-0.52	-0.4
Utah St.	12	S	0.33	0.76	1.14	1.54	1.92	2.2
Purdue		4	S	0.07	0.15	0.42	0.84	1.36	1.8
Siena		13	S	-0.07	-0.17	-0.15	-0.12	-0.12	-0.1
Notre Dame	6	S	-0.36	-0.55	-0.51	-0.48	-0.39	-0.3
Old Dominion	11	S	0.36	0.77	1.17	1.41	1.61	1.7
Baylor		3	S	-0.20	-0.55	-0.54	-0.56	0.06	0.8
Sam Houston St.	14	S	0.20	0.33	0.40	0.42	0.43	0.4
Richmond	7	S	-0.11	0.10	0.31	0.44	0.52	0.6
St. Mary's	10	S	0.11	0.51	0.92	1.18	1.38	1.5
Villanova	2	S	-0.06	-0.69	-1.82	-3.52	-3.95	-3.8
Robert Morris	15	S	0.06	0.07	0.07	0.06	0.06	0.1

What does this mean? Basically, you can't get value if you pick the usual suspects to win--if you pick them, somebody else already has and will probably get luckier. To maximize the likelihood of YOU winning the competition, you must adjust to take advantage of the herd mentality. So to maximize your likelihood of winning, you must take calculated risks. Basically, Wired Playbook explains the logic. Read it!

Obviously, if everyone has picked a team to go 2 rounds and then lose, if you pick them to win the title you gain on everyone else. A lot. If not, you'll lose badly.

If you pick the "consensus", you will finish about 80% but have no shot at winning.

 

[ChampsIn2011]

I really enjoy your statistical analysis.

It would be cool if you kept tally of your system's predictive power in the real world in an easily accessible way. Pretend to bet $10 on each game, as an example.

Also, running the numbers for previous seasons and comparing the results to the subsequent tournaments would be interesting.

 

[dyner]

hows the picks going so far?